用稳健统计法处理方法标准验证实验中方法检出限数据

方法检出限是分析方法基本性能参数之一。根据《环境监测分析方法标准制修订技术导则》(HJ 168—2010)的要求,此参数需要6家以上实验室基于特定样品测试数据的标准偏差得到。目前,标准文本大多采用各家实验室方法检出限中最大值。研究提出用稳健统计法处理方法验证数据,无需识别和删去离群值,可将离群值的影响降到最小。基于4组公开发表数据和2组实测数据,比较了不同方法计算结果。结果表明:参与统计的3组数据29个检测项目中,常规方法确认的方法检出限被判为离群值占27.6%。取最大值存在数值偏高的风险。提出了用曼德尔k检验法识别1组方法检出限数据中的离群值,该方法同样可以用于实验室数据与标准文本中方法检出限符合性判断。

Processing Data of Method Detection Limit in the Collaboration Experiment by Robust Statistics

Detection limit (DL) of method is one of the basic parameters related to analytical method performance. According to HJ 168-2010,this parameter can be obtained based on standard deviation data in a series of experiments following the same experimental program and the same number of repeated measurements by more than 6 laboratories. Most of standard text take the maximum one as DL in a series data reported by all laboratories. In this paper,the robust statistics (algorithm S) is used to process experiment data of method detection limit,which is not need to identify and delete outliers,but it can also reduce the effect of outliers on the final results based on all measurement data. Based on 4 sets of data in literatures and 2 set of experiment data collected by authors, four methods to calculate DL are compared. There is a high risk if the maximum standard deviation is chosen from all collected data as DL, which may be significantly higher than that expected. It is shown that the percent of the DL as outlier counting from 29 items in 3 sets of data is 27.6%. The rule of compliance judgment is proposed to determine whether the detection limit of method from own laboratory experimental data be consistent with that described in standard text, that is based on the ratio of two DL values being smaller or larger than critical value of Mandel k test. This ratio is also used to identify the outliers in a series of DL data.